A correction and analysis method for the spectral data collected by a specific infrared spectrum integrated system is presented. In the experiment, the infrared spectral data of a blackbody at a known specific temperature are firstly classified and sorted. The signal noises are eliminated as much as possible with different filtering algorithms and filter window widths on condition that the characteristic signals are maximally reserved. Then, the signals obtained are corrected by a partial least square algorithm so as to let them agree with the theoretical value. Finally, the correction results are used to construct a mathematical model and the concept of confidence is taken into account to verify the accuracy of the model. It is found in data analysis that the spectral data collected by the integrated infrared spectrum integrated system are close to the theoretical value after being corrected by the partial least-squares algorithm and the function relationship between the corrected series and the wave number can be fitted and is less affected by temperature.